Kevin Craib and colleagues1 report that “frequent [at least once daily] speedball (combined cocaine and heroin) injection” was a predictor of HIV seroconversion for male and female Aboriginal injection drug users in the Vancouver Injection Drug User Study (VIDUS). Other predictors, all identified by Cox regression analysis, were frequent cocaine injection, frequent heroin injection and self- reported “binges of injection drug use.” No other sociodemographic or behavioural characteristics were significant risk factors.
Needle-stick transmission of HIV is well known, so it is hardly surprising that high frequency of cocaine or heroin injection is statistically associated with increased rates of HIV infection. Moreover, 124 VIDUS participants died in the first approximately 5 years of the study, the cause of death being HIV/AIDS for 28, overdose for 41 and some other cause for 55.2 Thus, regardless of HIV transmission, frequent cocaine and heroin injection should be a public health concern, as well as a law enforcement issue.
The VIDUS investigators discerned previously3 that frequent speedball injection is significantly associated with a doubling of HIV seroconversion among all men and all women who enrolled as HIV-negative VIDUS participants. Yet Craib and colleagues1 assert that “The strong association between frequent speedball injection and HIV infection among both the Aboriginal men and women appears to be a new finding and is of grave concern.” They further state that their study “highlights the urgent need for the rapid implementation of evidence-based prevention interventions that are planned and delivered in partnership with Aboriginal AIDS service organizations and the Aboriginal community.” This VIDUS tune has an “aggressively assertive,” “presumptuous” and “overbearing” tone, which Sackett4 described as “the arrogance of preventive medicine.”
Craib and colleagues,1 using Kaplan–Meier plots of seroconversions over time (from individuals' dates of enrolment with HIV-negative status), show that cumulative HIV infection rates to 42 months after enrolment were twice as great for Aboriginal men and women as for non-Aboriginals. That is, Craib and colleagues repeat the finding of an earlier VIDUS paper,3 that Aboriginal identity is itself a risk factor for seroconversion. This factor is one that cannot be modified, and the medical literature already associates Aboriginal status with numerous health woes. Thus, it is difficult to understand the specific benefits of focusing on speedball injection.
In addition, Craib and colleagues1 disregard inflections and crossings of Aboriginal and non-Aboriginal time-to-event plots, which would suggest that the Cox proportional hazard regression model is not appropriate for these data.5 Nonetheless, this mathematical model drives all of the numeric, nongraphic risk estimations (and confidence intervals) presented by Craib and colleagues.1 In addition, they use log-rank tests to demonstrate that some of the differences shown graphically were statistically significant, but this point might be better demonstrated by 2 х 2 tables (counts of HIV-negative VIDUS enrollees cross-tabulated by race and by HIV status at 42 months, if known).
Good intentions and computing software can produce epidemiologic research that is highly “statisticated” but that is neither perceptive nor helpful.
Ned Glick Emeritus Professor Departments of Statistics and of Health Care and Epidemiology University of British Columbia Vancouver, BC